print("Restoring latest model from {}".format(CHECKPOINTS_PATH_SRM))
saver = tf.train.Saver()
latest_checkpoint = tf.train.latest_checkpoint(CHECKPOINTS_PATH_SRM)
print("Latest checkpoint: {}\n".format(latest_checkpoint))
saver.restore(srm_sess, latest_checkpoint)
#sys.exit(0)
if USE_SCORER and SCORER == "DCGAN_Scorer":
scorer_graph = tf.Graph()
scorer_sess = tf.Session(graph=scorer_graph, config=config)
with scorer_graph.as_default():
# data
galaxy_im, _, nb_galaxies, _ = create_dataloader_train_labeled(data_root=DATA_ROOT, batch_size=1, batches_to_prefetch=2, all_data=False)
# DCGAN Scorer model
print("Building DCGAN Scorer model ...")
sys.stdout.flush()
im_pl = tf.placeholder(dtype=tf.float32, shape=[BATCH_SIZE, 1, 1000, 1000])
model1 = DCGAN()
_, ops = model1.discriminator_model(inp=im_pl, training=False, resize=True) # get discriminator output
flat = ops["flat"]
model2 = Scorer_head()
scores_pred = model2.scorer_head_model(features=flat, training=False)
print("Restoring latest model from {}".format(CHECKPOINTS_PATH_SCORER))
saver = tf.train.Saver()
latest_checkpoint = tf.train.latest_checkpoint(CHECKPOINTS_PATH_SCORER)
create_zip_code_files(os.path.join(LOG_DIR, "code.zip"), files)
#sys.exit(0)
# remove warning messages
os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2"
tf.logging.set_verbosity(tf.logging.ERROR)
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
config.gpu_options.visible_device_list = "0"
with tf.Session(config=config) as sess:
# data
real_im, _, nb_reals, _ = create_dataloader_train_labeled(
data_root=DATA_ROOT,
batch_size=BATCH_SIZE,
batches_to_prefetch=BATCHES_TO_PREFETCH,
all_data=False)
training_pl = tf.placeholder(dtype=tf.bool, shape=[])
im_pl = tf.placeholder(dtype=tf.float32, shape=[BATCH_SIZE, C, H, W])
real_im = (real_im +
1) / 2.0 # denormalize images to the range [0.0, 255.0]
#model
print("Building model ...")
sys.stdout.flush()
model = CVAE(batch_size=BATCH_SIZE, latent_dim=NOISE_DIM)
max_pooled, mean, logvar, _ = model.inference_model(inp=im_pl,
training=training_pl)
print(" LOG_DIR: {}".format(LOG_DIR))
print(" CONTINUE_TRAINING: {}".format(CONTINUE_TRAINING))
print("\n")
sys.stdout.flush()
# remove warning messages
os.environ["TF_CPP_MIN_LOG_LEVEL"]="2"
tf.logging.set_verbosity(tf.logging.ERROR)
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
config.gpu_options.visible_device_list = "0"
with tf.Session(config=config) as sess:
# data
train_ds, nb_images = create_dataloader_train_labeled(data_root=DATA_ROOT, batch_size=BATCH_SIZE, batches_to_prefetch=BATCHES_TO_PREFETCH, all_data=True)
real_im, label = train_ds # unzip
# define noise and test data
noise = tf.random.normal([BATCH_SIZE, NOISE_DIM], seed=global_seed, name="random_noise") # noise fed to generator
y_G = tf.random.uniform( shape=[BATCH_SIZE, 1], minval=0, maxval=2, dtype=tf.int32, seed=global_seed) # labels fed to generator
noise_test = np.random.normal(0, 1, [BATCH_SIZE, NOISE_DIM]).astype("float32") # constant noise to see its evolution over time
y_test = np.random.randint(2, size=[BATCH_SIZE, 1]) # constant list of labels for testing
# define placeholders
im_pl = tf.placeholder(dtype=tf.float32, shape=[BATCH_SIZE, C, H, W]) # placeholder for real images fed to discriminator
noise_pl = tf.placeholder(dtype=tf.float32, shape=[BATCH_SIZE, NOISE_DIM]) # placeholder for noise fed to generator
y_pl_G = tf.placeholder(dtype=tf.float32, shape=[BATCH_SIZE, 1]) # placeholder for label fed to generator
y_pl_D = tf.placeholder(dtype=tf.float32, shape=[BATCH_SIZE, 1]) # placeholder for label fed to discriminator